Ensemble based speaker verification using adapted score fusion in noisy reverberant environments

摘要

This paper proposes an ensemble based automatic speaker recognition (ASV) using adapted score fusion in noisy reverberant environment. It is well known that background noise and reverberation affect the performance of the ASV systems. Various techniques have been reported to improve the robustness against noise and reverberation, and an ensemble based method is one of the effective techniques in the noisy environment. The ensemble based method uses a combination of several weak learners to achieve higher performance than a single learner method. However, since the performance is depended on the fusion weights, the adequate weight estimation method is required. The proposed weight estimation method is based a supervised adaptation and the evolutionary update algorithm. The QUT-NOISE-SRE protocol, which has been published recently, is used for simulating the reverberation of the clean speech in our experiments. The experimental results report the characteristics of the QUT-NOISE-SRE protocol and the effectiveness of the proposed method in noisy reverberant environment.